Re: The Nanotech Rapture


Steve O'Hara-Smith wrote:
Tim Tyler <seemysig@xxxxxxxxxxxxxx> wrote:

Computing elements will be among the first usable nanotechnology -

That may well be true - it's certainly a hungry market. However
that's not enough to provide AI. Consider, a human brain is reckoned to have
around 10^11 neurons, PCs routinely come with 3-5 * 10^11 bytes of storage
(hard disc) and can easily accomodate 4 * 10^12 bytes of storage (40 bytes
per neuron). This implies that modeling a human brain (or building a neural
network of human brain complexity) is in reach of a network of dedicated
PCs [...].

Right - but slow. Sometime around 2020 we will have suitable hardware
in $500 game consoles.

It may well be in reach of a brain@home project. Howver it is not being
tried as far as I can tell, indeed as far as I can tell nobody has any
clue how or where to start.

Therein lies the big difference between AI and molecular
moanufacturing - there are people who have a clue how to start on the
latter.

I don't really see it. Expert systems are all over the place.
My computer can recognise speech pretty well - considering its
CPU is smaller than my thumbnail. We know lots about how to build
AI.

and by my reckoning we'll probably have AI before we get down there.
Whether we need it or not, we will have it - and it will make quite
a difference.

It looks to be in reach today in terms of hardware capability,
software is a big problem. This is why I suggest that it may not turn up on
schedule for nanotech development. Human scale AI is no longer a hardware
problem.

At the very least, computing power for AI has to become cost-
competitive
with human brainpower before it can be widely deployed. It might make
sense for search companies today to invest in AI - e.g. see the
PowerSet
deal - but we may have to wait a decade before it is affordable in
desktops.
So, hardware still seems like an issue to me.
--
__________
|im |yler http://timtyler.org/ tim@xxxxxxxxxxx Remove lock to
reply.

.

Relevant Pages

  • Re: Merging with machines inevitable, scientists say
    ... we probably won't have hardware to match the ... >>> human brain at a reasonable cost for at least another ... >> why hasn't evolution discovered this and done away with ... >> the unnecessary messy details? ...
    (comp.ai.philosophy)
  • Re: Artificial life likely in 3 to 10 years
    ... ``The hardware problem is more challenging. ... the human brain neuron by neuron. ... Machines are streets ahead in hardware reliability at ...
    (sci.bio.evolution)
  • Re: Merging with machines inevitable, scientists say
    ... their period of superiority is probably not that far off. ... we probably won't have hardware to match the human brain at ... single cell, ...
    (comp.ai.philosophy)
  • Re: another bizarre architecture
    ... experimentally determined that the human brain cannot accurately ... (without having the hardware to run it on, I had to decide whether to ... as marginal as it gets) was the real time loop of a TI DSP, ... separate entity from the computer) the screenful rule of "thumb". ...
    (comp.arch.embedded)
  • Re: Radical Design May Result in New Slant on Aircraft
    ... The plane would need massive ... computing power, akin to that of the human brain, to continuously ...
    (sci.military.naval)
Quantcast